Image processor API

This specification defines the HTTP API used by Mankai remote image processors. A user adds the server’s base URL in Settings → Image Processing. Mankai reads the processor metadata, shows the server-defined configuration fields, and sends each reader image to the server when the processor is enabled.

For setup and everyday use in the app, see Image processing.

Endpoints

Method	Path	Authentication	Purpose
`GET`	`/`	Never	Return processor metadata and configuration
`POST`	`/process`	Optional	Process one image
`POST`	`/auth/login`	Never	Sign in when authentication is enabled
`POST`	`/auth/refresh`	Never	Refresh an access token

The URL entered by the user is the base URL. For example, entering https://images.example.com/mankai makes the processing endpoint https://images.example.com/mankai/process.

Processor information

`GET /`

Response — 200 OK

interface ImageProcessorInfo {
  id: string // Stable server/processor identifier
  name?: string
  version?: string
  description?: string
  authors?: string[] // Default: []
  repository?: string
  authenticationEnabled?: boolean // Default: false
  configs?: Config[] // Default: []
}

type ConfigType =
  'text' | 'password' | 'number' | 'slider' | 'boolean' | 'select' | 'color'

interface Config {
  key: string // Unique within this processor
  name: string // User-facing label or localization key
  description?: string
  type: ConfigType
  defaultValue: string | number | boolean
  options?: string[] // Used by "select"
  min?: number // Used by "slider"
  max?: number // Used by "slider"
  step?: number // Used by "slider"
  supportsOpacity?: boolean // Used by "color", defaults to false
}

Example:

{
  "id": "com.example.manga-denoise",
  "name": "Manga Denoise",
  "version": "1.0.0",
  "description": "Removes scan noise while preserving line art.",
  "authors": ["Example Lab"],
  "repository": "https://example.com/manga-denoise",
  "authenticationEnabled": false,
  "configs": [
    {
      "key": "strength",
      "name": "Strength",
      "description": "Higher values remove more noise.",
      "type": "slider",
      "defaultValue": 0.5,
      "min": 0,
      "max": 1,
      "step": 0.1
    },
    {
      "key": "preserveText",
      "name": "Preserve text",
      "type": "boolean",
      "defaultValue": true
    }
  ]
}

Image processing

`POST /process`

The request uses multipart/form-data so the image remains binary. It contains exactly these named parts:

Part	Content type	Value
`image`	`image/png`	The current pipeline image as a PNG file
`configs`	`application/json`	All current server-defined configuration values
`context`	`application/json`	The reader display size in logical points

configs uses the same key/value shape as the JavaScript plugin configuration API:

interface ConfigValue {
  key: string
  value: string | number | boolean
}

type ConfigValues = ConfigValue[]

Example configs part:

[
  { "key": "strength", "value": 0.8 },
  { "key": "preserveText", "value": true }
]

The context part has this shape:

interface ProcessContext {
  pointSize: {
    width: number
    height: number
  }
}

The server must return the processed image directly with a 2xx status and an image/* Content-Type. PNG, JPEG, WebP, HEIF, and any other image format supported by the client platform are allowed. The response is passed to the next configured image processor, so the server should preserve enough resolution and image quality for subsequent processing and reading.

Authentication (optional)

Set authenticationEnabled to true to make POST /process require JWT bearer authentication. Mankai then shows username and password fields and uses the same token flow as its HTTP plugins.

`POST /auth/login`

Request body

{
  "username": "user",
  "password": "secret"
}

Response — 200 OK

{
  "accessToken": "short-lived JWT",
  "refreshToken": "long-lived refresh token"
}

`POST /auth/refresh`

Request body

{
  "refreshToken": "long-lived refresh token"
}

Response — 200 OK

{
  "accessToken": "new short-lived JWT"
}